data types in csr matrix

[annonch]

Curiously using example code from the link in A2

The expected behavior differs from observed. Using int64 does not print the dtype information however int8 does.

Am I missing something?

>>> from scipy.sparse import *
>>> from scipy import *
>>> csr_matrix((3,4),dtype='int64').toarray()
array([[0, 0, 0, 0],
       [0, 0, 0, 0],
       [0, 0, 0, 0]])                                                            # Error
>>> csr_matrix((3,4),dtype='int8').toarray()
array([[0, 0, 0, 0],
       [0, 0, 0, 0],
       [0, 0, 0, 0]], dtype=int8)                                             # no error
>>> import numpy as np
>>> csr_matrix((3,4),dtype=np.int64).toarray()
array([[0, 0, 0, 0],
       [0, 0, 0, 0],
       [0, 0, 0, 0]])

[aronwc]

I wouldn't worry about the dtypes for this assignment. -Aron

…

On Wed, Oct 9, 2019 at 4:44 PM annonch ***@***.***> wrote: Curiously using example code from the link in A2 The expected behavior differs from observed. Using int64 does not print the dtype information however int8 does. Am I missing something? >>> from scipy.sparse import * >>> from scipy import * >>> csr_matrix((3,4),dtype='int64').toarray() array([[0, 0, 0, 0], [0, 0, 0, 0], [0, 0, 0, 0]]) # Error >>> csr_matrix((3,4),dtype='int8').toarray() array([[0, 0, 0, 0], [0, 0, 0, 0], [0, 0, 0, 0]], dtype=int8) # no error >>> import numpy as np >>> csr_matrix((3,4),dtype=np.int64).toarray() array([[0, 0, 0, 0], [0, 0, 0, 0], [0, 0, 0, 0]]) — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#493?email_source=notifications&email_token=AA575BPQS6K2QVHW7AOBS53QNZGCTA5CNFSM4I7FFJR2YY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4HQYLEJQ>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AA575BNTS3WSXXMWYHO7RKDQNZGCTANCNFSM4I7FFJRQ> .

[annonch]

Okay, just that my test for 'vectorize' will fail... My guess is platform dependency or version number.. Anyone else see this for osx? I use ports to install everything and pip for python packages.